To generate IGF2-tagged bsAbs, the synthetic gene encoding the mature IGF2 sequence (Δ1-7,F26S,R37K,R40K) was subcloned into the pMAZ-IgH and pMAZ-IgL vectors and linked at the N-terminus to the variable domain of the heavy and light chains of mAbs MR72 and mAb-548 via a short amino acid linker “ASTKGP” or “TVAAP”, respectively. NPC2-tagged bsAbs were similarly subcloned as above. To express these bsAbs, pMAZ-IgH and pMAZ-IgL encoding each antibody were co-transfected into Freestyle™ 293-F cells, (ThermoFisher Scientific) using linear polyethylenimine (Polysciences). Cells were incubated in Freestyle™ 293 expression media, at 37°C with 8% CO₂ in a shaking incubator for 6 days. Cells were pelleted and then the clarified supernatant was incubated with Protein A resin (1 ml packed resin per 600 ml clarified supernatant) for 2 h at 4°C. Antibodies were then purified using the gentle antibody elution system (ThermoFisher Scientific) per the manufacturer’s instructions. Eluted antibody was buffer-exchanged into Hepes buffer (200mM NaCl, 150mM HEPES[pH 7.4]) and concentrated using an Amicon centrifugal filter unit (Millipore Sigma) with a nominal molecular weight cutoff of 50 kDa.

Synthetic genes encoding C-terminally hexahistidine tagged CI-MPR domains 1–3 (amino acids 36–466) and domains 11–13 (amino acids 1508–1992) or NPC2 with a Twin-Strep-tag were subcloned into linearized pHL-sec vectors using AgeI and KpnI restriction sites. Proteins were expressed in Freestyle™ 293-F cells by transient transfection as above. Clarified supernatants containing CI-MPR domains were incubated with Ni-NTA resin (1ml packed resin per 300ml supernatant) at 4°C for 2 h prior to collection into a column. Resin was washed with wash buffer [500mM NaCl, 20mM HEPES(pH 7.6)] and protein eluted with elution buffer [500mM Imidazole, 500mM NaCl, 20mM HEPES(pH 7.6)]. NPC2 was purified from clarified supernatants using IBA Strep-Tactin^®; Sepharose^® (IBA) according to the manufacturer’s protocol. All proteins were dialyzed overnight in PBS and subsequently concentrated using an Amicon centrifugal filter unit (Millipore Sigma) with a nominal molecular weight cutoff of 10 kDa. Protein purity was assessed by SDS-polyacrylamide gel electrophoresis and visualized by staining with Coomassie Brilliant Blue G-250.

Soluble human NPC1 loop C was produced by IBT Bioservices as described previously (33). Ebola virus GPΔMuc [EBOV-Mayinga GP lacking the mucin-like domain (residues 312 to 462)] was expressed in Drosophila melanogaster S2 cells as previously described (14). EBOV GP_CL was produced by incubating trimeric GPΔMuc with thermolysin (Promega, V4001) in a 50:1 (protein:enzyme) ratio for approximately 18 h overnight at room temperature in 10 mM Tris-buffered saline [TBS; Tris-HCl(pH 7.5), 150 mM NaCl] containing 1 mM CaCl₂. Trimeric EBOV GP_CL was purified by Superdex S200 size exclusion chromatography in TBS immediately following incubation.

High-binding half-area 96-well ELISA plates (Corning) were incubated with 0.5 µg per well of CI-MPR domains 1–3 or 11–13 in PBS overnight at 4°C. Coated plates were blocked with blocking buffer (5% w/v nonfat dry milk in PBS) for 2 h at 37°C. A 3-fold dilution series of antibodies starting at 100 nM diluted in wash buffer (1% w/v nonfat dry milk in PBS) was incubated for 1 h at 37°C. Plates were washed 3 times with wash buffer. Anti-human IgG-HRP or anti-His tag-HRP secondary conjugates were diluted in wash buffer and incubated for 1 h at 37°C. Plates were washed with PBS and developed using Ultra-TMB (ThermoFisher) and quenched with 0.5M H₂SO₄. Absorbance was read at 450nm on a Cytation 5 cell imaging multi-mode reader (BioTek).

The OctetRed™ system (ForteBio, Pall LLC) was used to determine parental and bsAb binding properties. Anti-human Fc sensors were used for initial capture of IgG loading, and then measured for association to antigen at pH 5.5. Global fitting to a 1:1 binding model was used to estimate k_on (association rate constant), k_off (dissociation rate constant) and apparent K_D ^app (apparent equilibrium constant). Although data could be described accurately with a 1:1 model, given the bivalent nature of the antibody, and the trivalent nature of GP_CL, we cannot rule out avidity effects and therefore report apparent KD.

Human U2OS osteosarcoma cells were cultured in modified McCoy’s 5A medium (LifeTechnologies) supplemented with 10% fetal bovine serum (FBS; Atlanta Biologicals), 1% GlutaMAX (Life Technologies) and 1% penicillin/streptomycin (Life Technologies). Wildtype and cation-independent mannose-6-phosphate receptor (CI-MPR)–knockout human haploid Hap1 cells (Horizon Discovery, Cat# HZGHC006178c011) were cultured in Iscove’s modified Dulbecco’s medium (IMDM; ThermoFisher) and supplemented as above. Vero cells were maintained in high-glucose Dulbecco’s modified Eagle medium (DMEM; ThermoFisher) supplemented with 10% fetal bovine serum (FBS; Atlanta Biologicals), 1% GlutaMAX (Life Technologies) and 1% penicillin/streptomycin (Life Technologies). Human monocyte THP-1 cells (ATCC) were cultured in RPMI-1640 medium (ATCC, Cat #30-2001) supplemented with 10% FBS. All cells were maintained at 37°C, 5% CO₂ in a humidified incubator.

Generation and propagation of recombinant vesicular stomatitis viruses (rVSVs) encoding enhanced green fluorescent protein (eGFP) in the first position and replacing VSV G with glycoproteins from EBOV/Mayinga (EBOV/H.sap-tc/COD/76/Yambuku-Mayinga), TAFV (TAFV/H.sap-tc/CIV/94/CDC807212), BDBV (BDBV/H.sap/UGA/07/But-811250), BOMV (BOMV/Mops condylurus/SLE/2016/PREDICT_SLAB000156), SUDV/Boneface (SUDV/C.por-lab/SSD/76/Boneface), RESTV (RESTV/M.fas-tc/USA/89/Phi89-AZ-1435), LLOV (LLOV/M.sch-wt/ESP/03/Asturias-Bat86) and an rVSV encoding an mNeongreen-phosphoprotein P (mNG-P) fusion protein bearing MARV GP (MARV/H.sap-tc/KEN/80/Mt. Elgon-Musoke) were previously described (2, 5, 27, 34–37). An rVSV encoding eGFP in the first position and MLAV GP (MLAV/Rousettus-wt/CHN/2015/Sharen-Bat9447-1; GenBank Accession no: KX371887) was cloned and rescued as above (2, 5, 27, 38–41).

Dilution series of antibody or NPC2 were incubated on monolayers of human U2OS cells, wild type or CI-MPR–knockout human haploid Hap1 cells, or THP-1 cells for 2 h at 37°C prior to addition of pre-titrated amounts of rVSV-GP particles (MOI ~ 1 infectious unit (IU) per cell). For infection assays using THP-1 cells, cells were treated with 20ng/ml phorbol 12-myristate 13-acetate (PMA) (Millipore Sigma, Cat#P1585) for 72 h and media was exchanged prior to antibody incubation (42). Viral infectivities were measured by automated enumeration of eGFP⁺ or mNG⁺ cells using a Cytation 5 reader at 12-14 h post-infection. Data was subjected to non-linear regression analysis to extract half maximal inhibitory concentration (IC₅₀) values (4-parameter, variable slope sigmoidal dose-response equation; GraphPad Prism). Relative IC₅₀ values were calculated for all curves with sigmoidal curves and absolute IC₅₀ values were calculated for curves with ill-defined plateaus.

Parental antibodies and bsAbs were covalently labeled with pH-sensitive pHrodo Red succinimidyl ester (Thermo Fisher Scientific) according to the manufacturer’s instructions. Antibodies were incubated with 10-fold molar excess of pHrodo Red succinimidyl ester for 1 h at room temperature. Excess unconjugated dye was removed using PD-10 desalting columns (GE Healthcare). pHrodo Red-labeled antibodies were exchanged into HEPES buffer and concentrated in an Amicon Ultra centrifugal filter unit with a nominal molecular weight cutoff of 30 kDa. Antibody concentration and degree of labeling was determined according to the manufacturer’s instructions.

Pre-chilled confluent human U2OS cell monolayers were incubated with the pHrodo Red-labeled parental and bsAbs (50 nM) at either 4°C to prevent internalization or 37°C to allow internalization for 30 min. Cells were returned to ice and any unbound antibody was removed by washing with cold PBS. Cells were harvested using cold trypsin-EDTA for 10 min. Cells were washed with cold PBS prior to filtering through a mesh strainer. Single cells were analyzed for pHrodo Red fluorescence on a BD LSRII flow cytometer and FlowJo software.

A dilution series of antibodies were incubated on monolayers of U2OS cells for 14 h at 37°C. A 1:1 ratio of CellTiter-Glo reagent (Promega) to cell culture media was added per well. Contents were mixed for 2 minutes on an orbital shaker then incubated at room temperature for 10 minutes before luciferase activity was read using a Cytation 5 reader.

U2OS cells were seeded on fibronectin-coated coverslips. Cells were incubated with either media alone, U18666A (10 µM, Millipore Sigma), or antibody (1 nM or 350 nM) at 37°C for 14 h prior to fixing. Cells were washed with PBS prior to incubation with filipin III (Millipore Sigma) for 1 h at room temperature, washed again with PBS, and then mounted onto slides with Prolong (Thermo Fisher). Slides were imaged using a Zeiss Axio Observer inverted microscope with a 40x objective.

A dilution series of antibodies were incubated on monolayers of Vero E6 or U2OS cells for 2 h at 37°C prior to addition of pre-titrated amount of Ebola virus/H.sapienstc/COD/1995/Kikwit-9510621 (EBOV/Kik-9510621; ‘EBOV-Zaire 1995’) or Marburg virus/H.sapienstc/DEU/1967/Hesse-Ci67. At 48 h post-infection, cells were fixed with formalin, and blocked with 1% bovine serum albumin. EBOV-infected cells, MARV-infected cells and uninfected controls were incubated with either EBOV GP-specific mAb KZ52 or MARV GP-specific mAb 9G4, respectively. Cells were washed with PBS prior to incubation with either goat anti-human IgG or goat anti-mouse IgG conjugated to Alexa Fluor 488 (Invitrogen). Cells were counterstained with Hoechst stain (Invitrogen), washed with PBS and stored at 4°C. Infected cells were quantitated by fluorescence microscopy and automated image analysis using an Operetta high content device (Perkin Elmer) and the image analysis Harmony software, as previously described (27).

We explored a multifunctional cell-surface receptor, the cation-independent mannose-6-phosphate receptor/insulin-like growth factor 2 receptor (hereafter, CI-MPR), as an alternative endo/lysosomal delivery strategy for mAbs targeting the filovirus GP_CL:NPC1 interaction. CI-MPR is a ~300-kDa Type I membrane glycoprotein comprising a large extracellular domain with distinct binding sites for multiple ligands, a transmembrane region, and a cytoplasmic tail that regulates receptor internalization, endocytic trafficking, and recycling (43). Mannose-6-phosphate (M6P) and insulin-like growth factor 2 (IGF2) are the best characterized CI-MPR ligands and both interactions have been successfully exploited for endo/lysosomal delivery of recombinant cargo molecules (29–31, 44–46). Many lysosomal enzymes naturally undergo mannose-6-phosphorylation, affording their CI-MPR–mediated extracellular retrieval and intracellular delivery as enzyme replacement therapies for lysosomal storage disorders (34, 47–49). Here, we sought to exploit one such lysosome-resident protein, a 132-amino acid sterol-binding protein Niemann-Pick C2 (NPC2) (50, 51), to deliver mAb cargoes to late endosomes and lysosomes ( Figure 1A ). Accordingly, we fused NPC2 to mAbs MR72 and mAb-548 in two configurations (to the N–terminus of the IgG heavy or light chain) to create a panel of bsAbs. This panel was screened for neutralization potency against a surrogate vesicular stomatitis virus bearing EBOV GP (rVSV-EBOV GP) (39). We identified two highly active candidates mAb-548 and MR72 bearing NPC2 at the N–termini of their light chains (548~NPC2_LCN and MR72~NPC2_LCN, respectively; hereafter 548~NPC2 and MR72~NPC2) ( Figures 1B, C and Supplementary Figures S1A, B ).

To address concerns with inconsistent mannose-6-phosphorylation (M6P) of lysosomal proteins during manufacturing, LeBowitz and colleagues described a glycosylation-independent lysosomal targeting (GILT) approach in which recombinant cargoes were fused to the 61-amino acid mature IGF2 peptide sequence (30, 44). These proteins could be captured and internalized through IGF2:CI-MPR binding, thus bypassing the need for M6P. To test the efficacy of this approach for our mAbs, we generated a panel of IGF2-tagged bsAbs and down-selected them as above to identify two molecules bearing IGF2 at the N–termini of the mAb heavy chains (548~IGF2_HCN and MR72~IGF2_HCN, respectively; hereafter 548~IGF2 and MR72~IGF2) ( Figures 1B, C and Supplementary Figures S1A, B ).

We compared the dose-dependent antiviral activities of the down-selected bsAbs in the rVSV-EBOV GP infection assay in two cell types: U2OS human osteosarcoma cells and THP-1 human monocyte-like cells differentiated to macrophage-like cells with PMA. All four bsAbs afforded neutralization with low-nM IC₅₀ values and with ~10–100-fold greater potency than their parental IgGs, although the NPC2-tagged bsAbs were relatively more potent than their IGF2-tagged counterparts ( Figure 2 ). Importantly, only the bsAbs afforded complete neutralization. Consistent with their enhanced neutralizing activity, our top four bsAbs retained high-affinity binding to their respective endo/lysosomal target antigens at acid pH, as determined by biolayer interferometry (BLI) ( Figure 3 and Table 1 ). NPC2- and IGF2-tagged bsAbs displayed similar binding activities, indicating that the reduced activity of the latter is not a consequence of antigen:Ab binding penalties exacted by the NPC2 or IGF2 tag.

The CI-MPR ectodomain contains 15 ~150-amino acid ‘mannose 6-phosphate receptor homology’ domains (52). Domain 11 recognizes IGF2 (35, 53), and domains 3, 5, 9, and 15 recognize M6P (52) ( Figure 4A ). To begin to investigate the mechanisms of action of our Trojan horse bsAbs, we assessed their binding to recombinant, purified CI-MPR domains ( Supplementary Figures S1C, D ) by ELISA. Both IGF2-tagged bsAbs bound to a CI-MPR domain 11–13 construct predicted to recognize IGF2 but not M6P, whereas their NPC2-tagged and parental IgG counterparts did not ( Figures 4D, E ). Conversely, only 548~NPC2 bound to a CI-MPR domain 1–3 construct predicted to recognize M6P but not IGF2 ( Figure 4B ). Unexpectedly, MR72~NPC2 showed no detectable binding to domain 1–3 in the ELISA ( Figure 4C ), suggesting that the MR72 light chain may sterically hinder CI-MPR recognition by NPC2 but not by the much smaller IGF2 peptide. We hypothesize that reduced CI-MPR binding by MR72~NPC2 accounts for its ~10-fold reduced antiviral potency relative to 548~NPC2 ( Figures 1C and 2A, C, E, G ).

To test if the bsAbs require cellular CI-MPR for their antiviral activity, we next evaluated their capacity to neutralize rVSV-EBOV GP infection in isogenic haploid cell lines replete with or genetically deficient in CI-MPR (Hap1 and Hap1-CI-MPR^KO, respectively). Both NPC2-tagged bsAbs suffered an almost complete loss in activity in Hap1-CI-MPR^KO cells relative to the Hap1 cells, indicating that their antiviral activity was largely dependent on CI-MPR ( Figures 5A, C ). By contrast, the IGF2-tagged bsAbs were insensitive to CI-MPR loss ( Figures 5B, D ), indicating the existence of CI-MPR-independent pathways for their internalization into target cells (see below).

The basis of our strategy was that the NPC2- and IGF2-tags would mediate cellular receptor-mediated internalization and delivery of bsAbs to their endo/lysosomal sites of antiviral action. To evaluate this premise, we conjugated an acid-dependent fluorescent dye, pHrodo Red™, to each bsAb and its parent IgG, and measured the capacity of these conjugates to undergo temperature-dependent delivery to acidic intracellular compartments ( Figure 6A ), as described previously (27). As expected, no increases in single cell-associated pHrodo Red fluorescence was observed by flow cytometry with any of the mAbs or bsAbs following their incubation with cells for 30 min at 4°C. Similar results were obtained with the parent mAbs after 30 min at 37°C, as we observed previously (27). By contrast, considerable proportions of all four bsAbs were internalized under the same conditions, although the NPC2-tagged bsAbs appeared to undergo uptake more efficiently than did the IGF2-tagged molecules ( Figures 6B, C ). These findings support a model in which tag-mediated interaction with CI-MPR (NPC2) and/or other cell-surface receptors (IGF2) enhances bsAb internalization and delivery to NPC1-bearing compartments.

The superior cell-internalizing activity of the NPC2-tagged bsAbs relative to their IGF2-tagged counterparts suggested an explanation for the former’s enhanced antiviral potency. However, we also considered an additional hypothesis to explain these results. Specifically, NPC2 engages NPC1 as part of their interdependent function in cellular cholesterol trafficking and does so via a protein-protein interface that overlaps the filovirus GP_CL:NPC1 interface (54, 55). Thus, it remained possible that NPC2-tagged bsAbs compromise virus-receptor association directly through both mAb- and NPC2-dependent mechanisms. To address this possibility, we incubated cells with purified recombinant NPC2 ( Supplementary Figure S1E ), separately and in combination with mAb-548, and then exposed the cells to rVSV-EBOV GP. Neither treatment afforded a significant enhancement in viral neutralization relative to mAb-548 alone ( Figure 7 ). Further, only bsAbs bearing the NPC1-binding mAb-548 inhibited NPC1 function as measured by lysosomal accumulation of free cholesterol (56, 57), regardless of whether mAb-548 was fused to NPC2 or IGF2 ( Supplementary Figure S2A ). These results are consistent with a scenario in which NPC2 acts only to mediate endo/lysosomal delivery of the bsAbs and not as part of their antiviral payload.

Finally, we note that NPC1 inhibition was only observed following long-term treatment with high doses of bsAbs and did not compromise cell metabolic health or induce cytotoxicity ( Supplementary Figure S2B ).

The GP_CL:NPC1 interaction interface is highly conserved and required for all known mammal-infecting filoviruses, including both known and potential zoonotic threats (2, 5, 6, 19–22, 24). Given our evidence that the NPC2- and IGF2-tagged bsAbs could target this interaction to inhibit EBOV GP-dependent entry, we reasoned that they may possess broad anti-filovirus activity. Accordingly, we tested the bsAb panel against rVSVs bearing each of the filovirus GPs (ebolaviruses: BDBV, TAFV, BOMV, RESTV; SUDV, cuevavirus: LLOV, dianlovirus: MLAV, marburgvirus: MARV). All four bsAbs afforded substantial enhancements in neutralization potency relative to their parental IgGs against all of the rVSVs ( Figures 8A and Supplementary Figures S3, S4 ). The MR72-containing bsAbs alone provided only limited gains in neutralizing activity against rVSV-MARV GP ( Figures 8A , Supplementary Figures S3B, S4C, D ), likely reflecting MR72’s capacity to recognize and block the extracellular form of MARV GP but not of other filovirus GPs (27, 58). Finally, and importantly, these findings were concordant with the capacity of all four bsAbs to neutralize two divergent, authentic human disease-causing filoviruses, EBOV and MARV ( Figures 8B, C ). We conclude that the endo/lysosomal filovirus-receptor interaction represents a universal Achilles heel that can be targeted to develop pan-filovirus entry inhibitors.